Keyword "returns" in ANTLR4

Question

I'm currently adapting several ANTLR3 grammar files for ANTLR4. Some of the old ANTLR3 rules use the returns keyword. This rule is valid in both ANTLR3 and ANTLR4:

aRule:
  returns [String s]

  anotherRule AND (aRule)?
  {
      $s = "foo";
  }
;

...

Calling parser.aRule() on an ANTLR3 parser returns a String. Calling parser.aRule() on an ANTLR4 parser returns a RuleContext. Is the returns keyword in ANTLR4 purely for backward compatibility purposes? Obviously in both situations returns causes a public field String s to be added to the generated ARuleContext.

Would I be correct to assume ideally the returns keyword should no longer be used, and any sort of context state should be handled by a parse tree visitor?

To be honest, I'm not sure if the returns is still used/valid. In most cases I traverse and transform the parse tree manually, but a parse tree visitor may also do in cases where you don't need the extra flexibility provided by custom walking the parse tree. So in short: forget about the returns and do post processing in code afterwards based on the parse tree. — Onur

Bart Kiers Bart Kiers · Accepted Answer · 2014-09-16T19:50:27

Would I be correct to assume ideally the returns keyword should no longer be used, and any sort of context state should be handled by a parse tree visitor?

Yes, that is correct. With ANTLR4, you should keep all target specific code outside of your grammar and moved into listeners/visitors.

Keyword "returns" in ANTLR4

1 Answers